TensorFlow HVX Acceleration support

Question

I successfully built and ran the test application from https://github.com/tensorflow/tensorflow/tree/master/tensorflow/contrib/hvx. I'd now like to benchmark HVX against the CPU implementation of https://github.com/tensorflow/tensorflow/tree/master/tensorflow/tools/benchmark, and if possible, the Android camera demo, to see how much it would help, but I wasn't able to find any documentation describing how to build said apps with HVX support (my builds run on the CPU). I'm testing on the Open-Q 820 development board with Android 7.0.

Is utilizing HVX acceleration outside the HVX test application, preferably with the benchmark and maybe the Android camera demos supported yet? If so, could someone please point me in the right direction? Thanks!

Answer 1

Currently, the Android demo app does not support the HVX runtime. But I'm sure that you can use the runtime with Android demo app by replacing .so file with HVX version. If you can wait for the official support, that would be happening soon, but no promise. Let me know if you have any questions :)